Jorge
ProductHow it worksUse casesFAQ
Sign inJoin the waitlist
The context layer for the AI-native company

Your entire company.
One shared context.

Jorge knows how your company works. He gives every person and every AI agent the same structured knowledge, whether they use Claude, ChatGPT, Codex, Cursor, or whatever comes next.

Jorge meets new teams a few at a time, and he’ll reach out personally.
Usage-Based Billing — Northwind⌘K
SM
N
Northwind
Product
Files
PRDs
MDUsage-Based Billing.md
MDBilling Dashboard.md
RFCs
MDMetering Service.md
MDEvent Ingest.md
Architecture
MDOverview.md
MDData Pipeline.md
Decisions
MDDEC-022 Free Tier.md
Meetings
MDSep 03 Pricing Review.md
MDAug 31 Product Review.md
Personas
MDFinance Admin.md
Research
Attachments
PDFpricing-study.pdf
XLSusage-export.xlsx
DOCSales Playbook.docx
PPTQ3 Review.pptx
Finance
Synced~/Jorge/Northwind
MDUsage-Based Billing.md
MDMetering Service.md
Knowledge Graph
EditPreviewSplit
SMMCSaved 2m agoRev 14
Properties6 fields
Type
PRD
Status
Draft
Owner
SMSofia Mendes
Reviewers
Marina CostaRafael Lima
Related
RFC · Metering ServiceDEC-022
Updated
Sep 3, 2026 · 14:22
Usage-Based Billing
PRDOwned bySMSofia·Linked from 6 documents

Northwind charges every customer a flat monthly fee, whether they make 2,000 API calls or 2 million. Heavy users are underpriced, and small teams churn at renewal because the entry plan feels expensive. This PRD defines usage-based pricing and how it shows up in the Billing Dashboard.

Problem

In the Aug 31 review, sales reported that 7 of 10 deals lost in Q2 named price as the reason. The data is already there — the Metering Service counts every call — but invoices ignore it, so a 5-person team pays what a 500-person one does.

Goals
  • Free plan includes 10,000 API calls a month.
  • Bill usage above the plan per 1,000 calls, invoiced monthly.
  • Show customers their usage and projected bill before the invoice arrives.
Requirements
Plans: Free / Team / Business / Enterprise, each with a monthly call allowance.
Usage rolled up per workspace, with a 24 h grace window for spikes.
Credits and refunds need a reason; logged to Audit Trail.
Email the account owner at 80% of the allowance.
Open questions
  • Does the 24 h grace window hold for Event Ingest batch imports?
  • Persona check: does the Finance Admin need invoice exports in v1?
ChatHistory
What did the Sep 3 review decide about the grace window?
SM
Codex3 sources
Nothing final. The team agreed 24 h is "probably too generous for API spikes" and asked Rafael to check import batching. MDSep 03 Pricing Review

Two docs still assume 24 h: this PRD and MDMetering Service. Want me to add an open question to both?
Add to bothOnly this PRDNo
MDUsage-Based Billing.mdMDSep 03 Pricing Review.md
Ask about this workspace, or @mention a file…
CodexAsk before edits
SyncedAI readyCodex · gpt-5MarkdownUTF-8Ln 42, Col 8
Reads the same context
  • Claude
  • ChatGPT
  • Codex
  • Cursor
  • Gemini
  • MCP

Jorge in 50 seconds

Claude and Cursor shouldn’t disagree about your roadmap.

Jorge sits behind every assistant your team already uses. Switch tools and he brings the documents, decisions and citations with him.

Pick a question
Assistant
Claudeclaude.ai · MCPAsking Jorge · Northwind
How many API calls does the free plan include?

SourcesUsage-Based Billing.mdDEC-022 Free Tier.mdpricing-study.pdf p.14
Nothing is re-uploaded. Every assistant just asks Jorge.

Jorge reads your PRDs, RFCs and meeting notes and links them into one graph.

Your files stay plain files. Jorge reads them, links them and remembers who owns what, so nobody else has to.

Knowledge graph
Click a document to see what Jorge knows about it
PRD

Usage-Based Billing

Usage-based pricing for every plan. Target: the free plan includes 10,000 API calls a month.
6
connections
Sofia
owner
Connected to

Agents can’t turn your knowledge base into slop.

Every page has a type, and every type has a template. When an agent writes a page that breaks it, Jorge doesn’t save it. He tells the agent exactly what to fix. Free-form drafts go to a scratch area, out of everyone’s way.

ctx_writeRFCs/Overage Policy.mdCodex, signed in as Rafael Lima
What the agent sent
# Overage policy
So basically customers keep hitting the limit and
we should probably let them go over? A few
options here. Lots of thoughts below.
## Notes
- maybe 20% grace (see sales playbook?)
- talk to Marina
… 1,840 more words
What Jorge said
Waiting for the agent
Press step 1 to have Codex write a new RFC the way agents usually do.

Ask why the agent didn’t know. Jorge can tell you.

Agents get one call, ctx_resolve. Jorge returns the right sections, follows links one hop and stays inside the token budget. Every call is logged with what he returned, what he skipped, and why.

Each agent signs in as its person, gets that person’s access, and writes under their name.

ctx_resolve“How many API calls does the free plan include?”Budget
Token budget
1,950 / 2,000 tokensCodex as Rafael Lima · 14:22:07
  • Usage-Based Billing.mdGoals420Returned
  • DEC-022 Free Tier.mdDecision310Returned
  • pricing-study.pdfp.14 · §4.2 Free-plan limits680Returned
  • Sep 03 Pricing Review.mdDiscussion · linked from PRD540Returned
  • Metering Service.mdPlan table760Skipped · over budget
  • Aug 31 Product Review.mdMetrics900Skipped · over budget
  • DEC-014 Free Tier.mdDecision290Skipped · superseded
  • Finance/budget-2026.xlsxRevenue forecast350Skipped · no access

The PRD says 90 seconds. The decision record says 2 minutes.

Jorge rereads everything each morning. When two docs disagree, a link breaks or a page goes stale, he tells you before an agent repeats the wrong number to a customer.

78
Health score3 contradictions · 5 stale docs
Contradiction: free plan limit
MDUsage-Based Billing.md
PRD · Sofia · Sep 3
The free plan includes 10,000 API calls a month.
MDDEC-022 Free Tier.md
Decision · Marina · Aug 31
The free plan stays at 5,000 API calls a month.
The PRD raised the free limit to 10,000 after DEC-022 was accepted. Pick which one is true and Jorge will draft the fix to the other document for your review.

Keep your wiki for your team. Give your agents Jorge.

Your wiki is where your team writes. Jorge is where your team and its agents think together, organized by design and accountable by default.

Where Jorge wins
Docs & wiki toolsJorge
Built forPeople writing docs, with AI added laterAgents and people sharing one memory
StructureBlank page; anything goesTyped pages and templates, checked on every write
AI slopAgents pile up long, unstructured pagesSloppy writes are refused with a list of fixes
What agents getSearch, fetch, then sort through whole pagesOne call returns the right sections within a token budget
IdentityIntegration tokens, often sharedEach agent signs in as its person and writes under their name
“Why didn’t the agent know?”Can’t be answeredEvery retrieval is logged: returned, skipped and why
UpkeepManual gardeningBrain Health flags contradictions, stale and never-read pages
Working memoryNonectx_remember saves conversation summaries, secrets removed
Files for coding agentsExport or sync scriptjorge pull writes the workspace as markdown on disk
Lock-inProprietary block formatPlain markdown in and out
PricingPer seat, AI as an add-on tierPer person; AI assistants always free
Enforced templates and retrieval logs are the hardest for a blank-page tool to copy.

How Jorge helps each team

Team
  • PRDs

    Specs that stay true

    Every PRD links to the decisions, personas and research behind it, and Jorge tells you when one drifts.
  • Discovery

    Research you can ask

    Interviews, transcripts and decks become cited evidence your whole team can query.
  • Agents

    Drafts from real context

    Ask for a first draft; get one grounded in your own docs, staged for review.

Jorge is taking on new teams.

We bring on a few teams at a time and set Jorge up with you on a call. Tell him a little about your company and you’ll hear back within a week.

Role
Team size

Ask Jorge

Still curious? Write to hi@jorge.rocks. He answers.

Not a person, though he does rock. Jorge is the context layer your people and agents share: he reads your docs, keeps them linked and honest, and answers from them in whatever AI tool you use. That’s why he lives at jorge.rocks.

Claude, ChatGPT, Codex, Cursor and Gemini today, through native connectors and MCP. Anything that speaks MCP can read from Jorge, so new tools work on day one.

Your documents are plain files. Jorge can run local-first, with an encrypted sync for your team. Only the excerpts needed to answer a question are sent to the AI provider you choose, and every call is logged.

Agents stage edits; people apply them. You choose per folder what an agent can read, draft or create, and restricted folders stay invisible to agents run by people without access.

A folder of docs is enough: Markdown, PDF, Word, Excel or PowerPoint. Jorge detects PRDs, RFCs, decisions and meeting notes and links them automatically.

We onboard a few teams per cohort and set Jorge up with you on a call. Early teams get founder support and a say in the roadmap.

© Jorgehi@jorge.rocks